Developments on Microseismic Monitoring and Risk Assessment of Large-scale CO 2 Storage Bettina Goertz-Allmann NORSAR CCS Technical Workshop, RITE, Tokyo, January 16, 2019
Outline Introduction: Induced seismicity & CCS Two case studies In Salah, Algeria Event detection & analysis Confining event depth Decatur CCS site Microseismic event characterization Full-waveform modelling Summary & Conclusions
Induced Seismicity σ σ 3 - P p σ N [MPa] σ 1 - P p Increasing pore pressure (fluid injection) brings rock closer to failure Near-critical stress state (typical) microseismicity Examples: shale gas fracking, wastewater injection, geothermal stimulation, CO 2 injection
Induced Seismicity Example from Basel geothermal stimulation (Goertz-Allmann et al., 2011, 2012) Time-distance distribution of seismicity follows pressure diffusion Seismicity sometimes induced by very small pressure perturbations (tides, rain) At large perturbations (e.g. Basel), source parameters (stress drop and b-value) appear to correlate with pressure or differential stress.
Induced Seismicity Seismicity can be induced by a variety of other mechanisms: Depletion & compaction (e.g. oil production) Stress transfer in over- and sideburden Fracture opening (P > fracture pressure, «fracking») Fault reactivation Depletion & compaction injection
Induced Seismicity and CCS (Rutqvist, 2013) In CCS, we expect a number of the listed mechanisms to contribute to seismicity Pressure-induced seismicity mainly near injection and at plume front Stress-induced seismicity and other mechanisms further away Goals: 1. Verify seal integrity (risk assessment) 2. Track plume progression (monitoring) 3. Reservoir characterization (management / optimization)
The In Salah CO 2 storage site Injection commenced in 2004 via three injection wells.
The In Salah CO 2 storage site sandstones & mudstones carboniferous mudstones 20 m carboniferous standstone 4 MT CO 2 injected into a naturally fractured Carboniferous sandstone reservoir at 1.9 km depth.
Microseismic array at KB-601 2009-2011 Downhole array of 48 3C geophones between 30-500 m depth 6 geophones were connected to 3 digitizers GPS timing problems and strong electronic noise Only uppermost geophone provided reliable data
Goertz-Allmann et al. (2014) Event detection method Master event waveform crosscorrelation method to detect and pick seismic events within continuous data More than 5000 events are detected between August 2009 and June 2011
Comparison of events and injection data Goertz-Allmann et al. (2014) Kaiser effect? fracture pressure 155 MPa High correlation between occurrence of microseismic events and injection rate Periods of matrix injection and fracture injection
Event analysis using one geophone Differential S-P wave onset time gives an estimate of event-toreceiver distance Several clusters with similar arrival-time differences can be identified P S
Event analysis using one geophone Determine event direction from particle motion of P-waves. Futher separate clusters by combining S-P, azimuth, and inclination (clusters A-D). Overall events are oriented in the direction of the largest horizontal stress. No seismicity within a radius of about 1 km around the injection well -> Kaiser effect? (Oye and Ellsworth, 2005 BSSA)
Possible location of 4 event clusters Together with InSAR data the clusters may give an indication on the extent of the CO 2 plume in 2010. Goertz-Allmann et al. (2014)
Event analysis: example waveforms Additonal phase arrivals S-P ~ 1s S-P ~ 0.7s Shear-wave splitting stack S-P ~ 0.7s stack
M w estimation
M w estimation P-wave Distance and attenuation correction Determine the low-frequency spectral level Ω 0 (mean of 10 or 15 Hz to 20 Hz depending on SNR) Compute seismic moment m 0 and M W Most M W estimates are between -1 and 0, largest M W is about 1 M W P-wave > Mw S-wave Effect of the different radiation pattern S-wave
b-value analysis of event clusters b-value is the slope of the Gutenberg-Richter law b-value can be linked to in-situ reservoir stress state: e.g. high b-value when new fractures open and low b-value when pre-existing fractures are reactivated M C log10 (N) = a b M
b-value analysis of event clusters cluster A cluster B cluster C cluster D Similar b-values for P and S but significant variations between clusters. b ~ 1 for cluster A (average tectonic) Larger b (1.5 to 2) for clusters B-D
Shear-wave splitting analysis Up to 0.1 s Δt on S arrival. Sign of anisotropy time delay Eigenvalue method (Wüstefeld et al. 2010) gives anisotropy from Δt: A = (β Δt)/R stack A: percentage anisotropy β : average S velocity R : source receiver distance Δt: time delay
Shear-wave splitting analysis 83 events with good splitting 5% of anisotropy for clusters B-D Less than 2% of anisotropy for cluster A
Comparison to injection parameters No correlation between injection parameters and cluster A High correlation with clusters B-D High activity of cluster C only during main injection phase
Confining microseismic event depth P SP S ~ 0.5 s Additional phase on Z between direct P & S S-to-P converted phase at strongest velocity contrast (850 m). Cluster A
Confining microseismic event depth Use 3D ray tracing to identify converted SP. Test potential source locations: Waveforms at A and A have similar S-P traveltimes but converted phase only matches real data at shallower position A.
Confining microseismic event depth Cluster A at about 1.7 km (well above the reservoir but still within lower cap rock). No shear-wave splitting is observed and anisotropy may occur mainly in deeper layers. Inclination angles for cluster A are distinctly higher than for cluster B-D also pointing to a shallower source. Goertz-Allmann et al. (2014) Cluster A: within cap rock!
Summary of microseismicity at In Salah Over 5000 events detected despite only one sensor Clear dependence on CO 2 injection. Constrain event depth with later phase arrivals: we find that one cluster is in the cap rock. Two main groups of events are identified: Type 1: stress-triggered seismicity (A) Type 2: fracture opening (B-D)
Outline Introduction: Induced seismicity & CCS Two case studies In Salah, Algeria Event detection & analysis Confining event depth Decatur CCS site Microseismic event characterization Full-waveform modelling Summary & Conclusions
Inject 1 M tons of CO 2 into Mt. Simon sandstone (460 m thick) at about 1.9 km depth (end 2011-2014). Borehole & surface sensors. About 4,800 microseismic events were located using borehole strings. Events occur in distinct clusters with heterogeneous activity. The Decatur CCS site Mt. Simon Lower Mt. Simon Argenta Precambrian basement
The Decatur CCS site Most events with M w < 0. Injection at very low pressure (< 1 MPa) No obvious correlation with plume migration events far from the injection
Spatial distribution of event clusters and comparison to modelled CO 2 plume Map view EW cross section Cluster centroid locations Borehole stations Modeled CO 2 plume 16 microseismic clusters
Event characterization Basement Reservoir vs. events reservoir events events Vp [m/s] direct wave head wave Depth [m] Precambrian Theoretical ray diagrams for reservoir & basement events. Different waveform signature: head wave and direct wave arrivals clearly visible for reservoir events
Event characterization Reservoir Basement
Sub-cluster analysis of one cluster Event characterization Cross-correlation matrix Goertz-Allmann et al. (2017)
Microseismic event characterization Cluster A Spatial distribution Temporal migration Separate events occurring within different layers: Cold = reservoir Warm = basement Migration of events from the reservoir into the basement over the course of 100-200 days. Goertz-Allmann et al. (2017), JGR
Cluster B spatial Event characterization temporal b-value stress drop Separation between reservoir and basement events. Migration from the reservoir into the basement. Decrease of b-value with distance. Increase of stress drop with distance. Evidence for a fluid-driven process at the cluster level. Signs of pressure diffusion. Possible punctual hydraulic connection between reservoir and basement (i.e., confined to faults).
Relative event locations Preliminary results of improved relative event locations by developing a modified relocation method. Accurate event locations are necessary for any kind of interpretation Change of cluster orientation Planar feature Fracture? Old event locations Relocated events
PS3_2 PS3_1 Full-waveform modelling Observed waveform example Pn P Sn S Different phase arrivals with head wave and direct wave arrivals. Pn/Sn phase arrives first at deeper sensor (PS3_1). P/S phase arrives first at shallower sensor (PS3_2). Waveform modelling can help us to better understand the observed waveform characteristics. Gain a complete picture of the travel path of an event and helps us to select events and phases, which best sample the target area.
Full-waveform modelling 3D FD modelling using 1D velocity model 30 Hz Ricker wavelet. Compare sources placed at 1600 m, 2040 m, and 2200 m depth. Synthetic receivers Real receivers Sources at 1600 m, 2040 m and 2200 m
Full-waveform modelling Source at 2040 m depth
Full-waveform modelling Sequential snap-shots of full waveform modelling (from A to F)
Full-waveform modelling observed modelled
Full-waveform modelling Source 1600 m Source 2040 m modelled Different source depths show different signatures. Best match between observed and modelled data at 2040 m (reservoir/basement interface). Different phases can only be distinguished at larger source-receiver distances (> 1200 m). Source 2200 m observed
Summary of microseismicity at Decatur Type 1: mainly stress-triggered seismicity A seismicity migration pattern from the sediment into basement is observed. Seismicity within a cluster exhibits signs of pressure diffusion, both through the spatio-temporal evolution of seismicity but also through source parameters such as b-value and stress drop. Eventually, a punctual hydraulic connection (such as, e.g., a basement-connected fault) causes migration into the basement. may explain clustering of seismicity (i.e., weak crust around those areas). Finite-difference modelling helps to correctly identify seismic phases sampling the CO 2 plume and can confirm a source at the reservoir/basement interface.
Comparison In Salah and Decatur Depth resolution of microseismicity was critical to support reservoir characterization Obtained by exploiting information contained in later arrivals / multipathing Requires waveform modelling for hypothesis testing and confirmation In Salah: Information on caprock integrity Variations in b-value between (pressure-driven?) reservoir events and (stress-driven?) caprock events Despite very inadequate network coverage Decatur: Connection between reservoir and basement Overall stress-driven seismicity/ fracture reactivation But fluid-driven characteristics within cluster
More general insight During CCS operations: most important is event depth resolution to verify seal integrity. Reservoirs are generally thinner than depth uncertainty from standard seismological methods. Therefore, additional constraints need to be exploited to improve depth resolution. Integration of reservoir engineering data is important for meaningful microseismic interpretation: Need pressure, pumping & fluid flow data densely sampled in time with accurate time stamp. Source parameters (b, σ) can provide hints of reservoir hydraulics, but require good calibration Accurate moment magnitudes are important for risk assessment (event discrimination and forecast with b value): Ensure sufficient bandwidth of recording Good prior knowledge of noise environment (particularly problematic offshore)
More general insight Good network planning: Vertical aperture with borehole array(s) for depth resolution Azimuthal coverage for location accuracy and source parameter inversion including moment tensor Real-time data stream and automatic processing can provide traffic light feedback to operations. In most cases I have seen, microseismicity can NOT be used to track CO 2 plume because of often lack of brittle deformation. Therefore a complementary method needs to be used for that (4D seismic, InSAR (deserts!), microgravity (offshore), geochemical sampling,. ).
Contact: Bettina Goertz-Allmann Email: bettina@norsar.no Thank you for your attention! Thanks to my collegues: D. Kühn, V. Oye, K. Iranpour, B. Dando, E. Aker, B. Bohloli, R. Bauer, S. Greenberg, Acknowledgements The partners in the In Salah CO2 Storage Joint Industry Project - BP, Equinor (formerly Statoil), and Sonatrach. Part of this work on Decatur was supported as part of the Center of Geological Storage of CO 2, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science. Data for this project were provided, in part, by work supported by the U.S. Department of Energy under award number DE-FC26-05NT42588 and the Illinois Department of Commerce and Economic Opportunity. We are grateful for support through the Climit program of GASSNOVA project No. 616065. Data provided by the Midwest Geological Sequestration Consortium (MGSC), funded through the U.S. DoE National Energy Technology Laboratory (NETL) and the State of Illinois.